Messages

1111

General Questions and Answers / Re: Why am I getting a small bandgap in density of states of graphene?

« on: March 24, 2017, 15:33 »

You should shift the k-grid to the G point, and make sure that the grid chosen also includes the K-point, as this is the point where the graphene conduction and valence bands touch. Regarding DOS calculation for graphene, you may take a look at Appendix in Phys. Rev. B 87, 075414 (2013).

1112

General Questions and Answers / Re: Which one would be more suitable, LDA or GGA?

« on: March 22, 2017, 17:07 »

None of the local (LDA) and semi-local (GGA) density functionals account for the van der Waals interaction that determines separation distance between the monolayers in layered materials. Please  have a look at the following tutorial on how the van der Waals interaction correction can be taken into account in the ATK-DFT calculations, see http://docs.quantumwise.com/tutorials/dispersion_corrections_and_bsse/dispersion_corrections_and_bsse.html 

1113

General Questions and Answers / Re: Optimize Geometry

« on: March 22, 2017, 16:14 »

1. This is the maximum step length the geometry optimizer may take. If you have a problem converging to equilibrium geometry during optimization, you may want to reduce the maximum step length.

2. You may optimize all the lattice parameters simultaneously, or any pair of them, or a particular one, or none of them, depending on how you set constraints in the geometry optimization settings (Constrain Lattice Vectors). The fault is not to optimize any of the lattice parameters, i.e., to do ion relaxation only.

1114

General Questions and Answers / Re: converge problem all again.

« on: March 21, 2017, 11:40 »

To understand the issue related to finite bias calculation for perfect systems, you may want to consult a reference book on Electron Transport by S. Datta, http://www.cambridge.org/catalogue/catalogue.asp?isbn=0521599431, e.g., Chapter 2.1.

In brief, you make an assumption in your calculations that the potential drop takes place entirely across a perfect conductor. But this assumption is only valid if the number of conducting channels in the electrodes is infinite or at least much larger than the number of conducting channels in the conductor. That would be true if you would have connected your perfect conductor to metallic leads, but in your case the leads and conductor are made of the same perfect material. It means that the number of channels is the same in the leads and conductor, in contradiction with your assumption that the potential drop occurs essentially across the conductor, which is given by the central region in your ATK transport calculations.

That is why I said that your finite-bias calculations do not make sense for the device setup assumed.     

 

1115

General Questions and Answers / Re: energy independent kappa value in Complex band structure of HfO2

« on: March 20, 2017, 17:14 »

I guess the kappa still depends on the energy, but the dependence is just rather weak. To see a connection of the corresponding imaginary bands to real bands you may need to plot the band structure over much larger energy range.

The difference between the HfO2 and WSe2 complex band structures is very likely due to a larger band gap of HfO2 compared to that of WSe2, meaning that the HfO2 imaginary bands (or in other words, evanescent states) are more localized in some way.

I would like to notice that these states are not real physical states in a perfect bulk material such as HfO2.  The evanescent states may however couple, for example, to metal electrode states if a HfO2 dielectric spacer is sandwiched between two metal electrodes. To understand this, you may think of a simple 1D barrier problem, when plane-wave states are matched to exponentially-decaying states at the two barrier edges, for a given energy (which is below the top of the barrier indeed). 

1116

General Questions and Answers / Re: how do we calculate spin transport in finite bias?

« on: March 20, 2017, 13:06 »

Yes, you are right.

1117

General Questions and Answers / Re: converge problem all again.

« on: March 19, 2017, 23:08 »

It seems that your density mesh cut-off is rather small; default is 75 H, whereas you have adopted 36 H. I also notice that your system has no scatters in the central region, so I am not sure that doing finite bias calculations makes much sense in this case of a perfect system. The linear response calculation (when V_bias = 0 V) of conductance is what describes the electron transport through a periodic structure of a perfect, defect-free material like yours.

1118

General Questions and Answers / Re: Static dielectric constant and High Frequency dielectric constant

« on: March 19, 2017, 11:48 »

Do you mean how these can be experimentally measured? If so, you may google this kind of information, e.g., see https://www.google.dk/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#q=measure+dielectric+constant&*.

1119

General Questions and Answers / Re: converge problem all again.

« on: March 19, 2017, 11:41 »

Did you use the self-consistent solution of your zero-bias device calculation to start the finite-bias device calculation, see the tutorial http://docs.quantumwise.com/tutorials/atk_transport_calculations/atk_transport_calculations.html? You may also try increasing the bias in smaller steps, reusing a previous calculation for the next one. 

I notice that the script enclosed seems to be incomplete, as there is information about the device geometry only, no any other potentially useful information that might give more clues on what might have gone wrong with your device calculation.

1120

General Questions and Answers / Re: Bloch state velocity in silicene = 0 !!!

« on: March 14, 2017, 21:39 »

The conical point in the graphene's band structure is peculiar as I was saying. If you calculate the velocity exactly at k=[1/3, 1/3, 0], without the shift of 0.001 assumed in the original script, you will actually see that the V_y component value is close to that of V_x, i.e., it is of the same order of magnitude. So, the velocity components seem to be sensitive to the shift value. 

1121

General Questions and Answers / Re: Bloch state velocity in silicene = 0 !!!

« on: March 14, 2017, 18:31 »

The velocity components are defined as follows: velocity[0][0][0]  ->  V_x, velocity[0][0][1]   ->  V_y, velocity[0][0][2]  ->  V_z.  For the nanoribbon, you should take the velocity component that is along the nanoribbon; if the confinement directions are X and Y, then this will be V_z=velocity[0][0][2]. Note that the first index goes over all the bands specified in band_indices. If no band_indices are specified, the velocity will then be calculated for all the bands at a given k-point, e.g., velocity[0][0][0],  velocity[1][0][0] and so on for the X component of the first, second and so on bands, respectively.

Regarding graphene, V_z = 0 as it should be for the out-of-plane (z) direction. V_x and V_y are roughly the same if one calculates the velocity exactly at k=[1/3, 1/3, 0]. In principle, they are not expected to be identical because of band structure warping, see Appendix in PHYSICAL REVIEW B 87, 075414 (2013), i.e., the commonly assumed dispersion (E~V |k|) of graphene bands near the conical point is an approximation. The velocity is however somewhat sensitive to the k-point shift due to numerical issues I guess, as the conical point is a peculiar point in the graphene band structure.  Personally, I would calculate the graphene's velocity at the Dirac point from the density of states (using a very dense k-mesh as discussed in the PRB paper mentioned), assuming that E~V |k|, unless one really wants to study the effect of warping on the graphene's band structure. 

1122

General Questions and Answers / Re: Sabalcore Problem

« on: March 14, 2017, 16:27 »

If you are still having the issue with connecting to Sabalcore, you may try going through the following tutorial: http://docs.quantumwise.com/tutorials/job_manager_remote/job_manager_remote.html
and especially this one: http://docs.quantumwise.com/tutorials/ssh_keys/ssh_keys.html#ssh-keys.

1123

General Questions and Answers / Re: self consistent with SOC

« on: March 14, 2017, 16:04 »

I guess you can do it in this way indeed, provided that the spin type is consistently used through all these computational steps, as you have actually done, and this is why it worked for you.

1124

General Questions and Answers / Re: Bloch state velocity in silicene = 0 !!!

« on: March 14, 2017, 15:46 »

This can be any point between G and Z k-points, i.e., k=[0,0,kz], depending where you want it to be evaluated. You also have to decide on the band (given by band_indices in the script) for which you actually want to calculate the Bloch-state velocity. The velocity component along the nanoribbon (Z direction) should then be calculated. I note that the components along the X and Y are zero since the electrons are confined within the nanoribbon in these two directions, meaning that, e.g., the velocity component along the X direction, velocity[0][0][0], calculated in the script is zero for your system. 

1125

General Questions and Answers / Re: self consistent with SOC

« on: March 12, 2017, 21:08 »

Could you enclose the python script related to your calculation? What is the ATK version you are using?